Variant caller assignment comparison for Ion Torrent data

YANINA MURUA; CRISTOBAL FRESNO; MARCELA MENA; JUAN MARTÍN SENDOYA; OSVALDO PODHAJCER; ANDREA LLERA; ELMER FERNÁNDEZ

Conferencia; 4th International Society for Comutational Biology-Latin America Bioinformatics Conference (ISCB-LA); 2016

Background: Ion Torrent NGS sequencer users perform the variant calling process with Torrent Variant Caller (TVC) and/or Ion Reporter (IR), both provided  free-of-charge for academic users.IR is a proprietary suite which may not be customized. In order to explore the possibility of other available open source variant analysis tools which can be used for different NGS technologies, we compared different germline variant calling pipelines on variant analysis of a customized Ampliseq Ion Torrent panel.Methods: Four candidate genes related to a genetic disease were selected and tested with 167 amplicons at an average of 800X coverage. Nine pipelines of variant analyses were tested that differed in the use of  raw/trimmed reads, aligners (BWA-MEM and TMAP), base quality recalibration and variant callers - HaplotypeCaller (HC),  VarScan (VS), and compared with results obtained by TVC and IR.Results: The lack of base recalibration generated a huge number of false positive variants for HC and VS. On the contrary, IR reported the lowest number of variants of all tested callers. Read trimming did not affect the performance significantly. Together, recalibrated HC, recalibrated VS and TVC analyses had similar performances in some aspects but failed to detect ⅓ of IR variants. Conclusions: Germline variant discovery should be carried out using a combination of base quality recalibrated data and selected variant calling pipelines in order to avoid losing biological insight of the problem at hand. Further analysis is now focused in validating variants obtained by single pipelines by use of annotation tools and Sanger sequencing.